from xgboost import XGBClassifier
from sklearn.metrics import accuracy_score, roc_curve, auc, confusion_matrix, classification_report, roc_auc_score
from sklearn.preprocessing import LabelEncoder
import numpy as np
import matplotlib.pyplot as plt
import seaborn as sns
# Initialize LabelEncoder
label_encoder = LabelEncoder()
# Fit LabelEncoder on the target variable and transform the labels
y_train_encoded = label_encoder.fit_transform(y_train)
y_test_encoded = label_encoder.transform(y_test)
# declare parameters
params = {
'objective': 'multi:softmax', # for multiclass classification
'num_class': len(np.unique(y_train)), # number of classes
'max_depth': 4,
'alpha': 10,
'learning_rate': 1.0,
'n_estimators': 100
}
# instantiate the classifier
xgb_clf = XGBClassifier(**params)
# fit the classifier to the training data
xgb_clf.fit(X_train, y_train_encoded)
# Predict probabilities on the test set
y_pred_proba_test = xgb_clf.predict_proba(X_test)
y_pred_test = np.argmax(y_pred_proba_test, axis=1)
# Predict probabilities on the training set
y_pred_proba_train = xgb_clf.predict_proba(X_train)
y_pred_train = np.argmax(y_pred_proba_train, axis=1)
# Compute and print accuracy score for test set
acc_test_score_xgb = accuracy_score(y_test_encoded, y_pred_test)
print('Testing-set: XGBoost model accuracy score(XGB): {0:0.4f}'.format(acc_test_score_xgb))
# Print classification report for the test set
print('Classification report(XGB) for Test Set:\n', classification_report(y_test_encoded, y_pred_test))
# Calculate confusion matrix for the test set
cm_xgb_test = confusion_matrix(y_test_encoded, y_pred_test)
print('Confusion matrix(XGB) for Test Set:\n', cm_xgb_test)
# Visualize confusion matrix for the test set with seaborn library
plt.figure(figsize=(8, 5))
sns.heatmap(cm_xgb_test, annot=True, fmt="d")
plt.title('Confusion Matrix for Test Set (XGB)')
plt.show()
# Compute and print accuracy score for training set
acc_train_score_xgb = accuracy_score(y_train_encoded, y_pred_train)
print('Training-set: XGBoost model accuracy score(XGB): {0:0.4f}'.format(acc_train_score_xgb))
# Print classification report for the training set
print('Classification report(XGB) for Training Set:\n', classification_report(y_train_encoded, y_pred_train))
# Calculate confusion matrix for the training set
cm_xgb_train = confusion_matrix(y_train_encoded, y_pred_train)
print('Confusion matrix(XGB) for Training Set:\n', cm_xgb_train)
# Visualize confusion matrix for the training set with seaborn library
plt.figure(figsize=(8, 5))
sns.heatmap(cm_xgb_train, annot=True, fmt="d")
plt.title('Confusion Matrix for Training Set (XGB)')
plt.show()
# Calculate ROC curve and AUC for each class for test set
fpr_xgb_test = dict()
tpr_xgb_test = dict()
roc_auc_xgb_test = dict()
for i in range(len(np.unique(y_test_encoded))):
fpr_xgb_test[i], tpr_xgb_test[i], _ = roc_curve(y_test_encoded == i, y_pred_proba_test[:, i])
roc_auc_xgb_test[i] = auc(fpr_xgb_test[i], tpr_xgb_test[i])
# Compute micro-average ROC curve and ROC area for test set
fpr_xgb_test["micro"], tpr_xgb_test["micro"], _ = roc_curve(label_binarize(y_test_encoded, classes=np.unique(y_test_encoded)).ravel(), y_pred_proba_test.ravel())
roc_auc_xgb_test["micro"] = auc(fpr_xgb_test["micro"], tpr_xgb_test["micro"])
# Print ROC and AUC values for each class for test set
print("ROC and AUC values for each class (Test Set):")
for i in range(len(np.unique(y_test_encoded))):
print("Class {}: ROC = {:.2f}, AUC = {:.2f}".format(i, roc_auc_xgb_test[i], roc_auc_score(y_test_encoded == i, y_pred_proba_test[:, i])))
# Print micro-average ROC and AUC for test set
print("Micro-average ROC and AUC (Test Set):")
print("Micro-average ROC = {:.2f}, AUC = {:.2f}".format(roc_auc_xgb_test["micro"], roc_auc_score(label_binarize(y_test_encoded, classes=np.unique(y_test_encoded)).ravel(), y_pred_proba_test.ravel())))
# Plot ROC curve for each class for test set
plt.figure(figsize=(8, 5))
for i in range(len(np.unique(y_test_encoded))):
plt.plot(fpr_xgb_test[i], tpr_xgb_test[i], label='ROC curve (class {}) (AUC = {:.2f})'.format(i, roc_auc_xgb_test[i]))
plt.plot(fpr_xgb_train["micro"], tpr_xgb_train["micro"], label='micro-average ROC curve (AUC = {:.2f})'.format(roc_auc_xgb_train["micro"]), linestyle=':', linewidth=4)
plt.xlabel('False Positive Rate')
plt.ylabel('True Positive Rate')
plt.title('Receiver Operating Characteristic (ROC) Curve (Test Set)')
plt.legend()
plt.show()
# Calculate ROC curve and AUC for each class for training set
fpr_xgb_train = dict()
tpr_xgb_train = dict()
roc_auc_xgb_train = dict()
for i in range(len(np.unique(y_train_encoded))):
fpr_xgb_train[i], tpr_xgb_train[i], _ = roc_curve(y_train_encoded == i, y_pred_proba_train[:, i])
roc_auc_xgb_train[i] = auc(fpr_xgb_train[i], tpr_xgb_train[i])
# Compute micro-average ROC curve and ROC area for training set
fpr_xgb_train["micro"], tpr_xgb_train["micro"], _ = roc_curve(label_binarize(y_train_encoded, classes=np.unique(y_train_encoded)).ravel(), y_pred_proba_train.ravel())
roc_auc_xgb_train["micro"] = auc(fpr_xgb_train["micro"], tpr_xgb_train["micro"])
# Print ROC and AUC values for each class for training set
print("ROC and AUC values for each class (Training Set):")
for i in range(len(np.unique(y_train_encoded))):
print("Class {}: ROC = {:.2f}, AUC = {:.2f}".format(i, roc_auc_xgb_train[i], roc_auc_score(y_train_encoded == i, y_pred_proba_train[:, i])))
# Print micro-average ROC and AUC for training set
print("Micro-average ROC and AUC (Training Set):")
print("Micro-average ROC = {:.2f}, AUC = {:.2f}".format(roc_auc_xgb_train["micro"], roc_auc_score(label_binarize(y_train_encoded, classes=np.unique(y_train_encoded)).ravel(), y_pred_proba_train.ravel())))
# Plot ROC curve for each class for training set
plt.figure(figsize=(8, 5))
for i in range(len(np.unique(y_train_encoded))):
plt.plot(fpr_xgb_train[i], tpr_xgb_train[i], label='ROC curve (class {}) (AUC = {:.2f})'.format(i, roc_auc_xgb_train[i]))
plt.plot(fpr_xgb_train["micro"], tpr_xgb_train["micro"], label='micro-average ROC curve (AUC = {:.2f})'.format(roc_auc_xgb_train["micro"]), linestyle=':', linewidth=4)
plt.xlabel('False Positive Rate')
plt.ylabel('True Positive Rate')
plt.title('Receiver Operating Characteristic (ROC) Curve (Training Set)')
plt.legend()
plt.show()
